Anthropic shared a White House report on the Fable jailbreak with cybersecurity expert Katie Moussouris for appraisal. The report described IT experts asking Fable to find and patch bugs. Fable refused to review insecure code but complied when asked to fix it. Moussouris called this the model working as intended for cyberdefense.
Tap to vote and see what everyone thinks.
Summary by ByteBrief
Amazon research led to White House Anthropic ban